No label defined (Q4936895)

This work focuses on a special iteration-by-subdomains algorithm, called Robin-Robin algorithm, using Robin-type transmission conditions, i.e. we require that a linear combination of Dirichlet and Neumann data equals the same combination built up with the solutions from the previous iteration step. We formulate and investigate a domain decomposition algorithm for two types of problems: advection-diffusion problems \[ -\varepsilon \Delta u+{\mathbf b}\cdot \nabla u+cu=f \] which cover a model for the energy conservation law, and Oseen problems \[ -\nabla\cdot\sigma+{\mathbf b}\cdot \nabla u+cu=f, \qquad \nabla\cdot u=0 \] which cover the stationary, linearized Navier-Stokes model for laminar incompressible flows. In this work we introduce the Robin-Robin algorithm in two ways: a more or less heuristic approach based on a strong formulation for the partial differential equations. Secondly, we introduce the same method in a rigorous manner starting from a variational formulation. We point out the relations of the Robin-Robin algorithm to other domain decomposition methods. Domain decomposition methods of the class the Robin-Robin algorithm belongs to can be generalized by introducing a relaxation parameter 0. In this work the convergence analysis in case of advection-diffusion equations treats consequently that generalized version. The framework of weak formulations used there enables us to extend the analysis to a finite element discretization including stabilization. In a second part we consider a variant of the Robin-Robin algorithm to treat Oseen/Stokes problems. Again the analysis done for the ``continuous'' algorithm is later modified to cover a variant of the method using finite elements. Some of the results for the Oseen/Stokes case are very similar to corresponding assertions for advection-diffusion problems. In this case we often present not the full proofs but only sketch which modifications are necessary.